Renewable energy integration Some examples:

DC/AC
ZF Grid
Single-stage PV system: ZT ~
C

DC/AC
ZF Grid
Two-stage PV system: DC/DC
ZT ~
C

Doubly-Fed Induction Permanent Magnet

Generator (DFIG): Synchronous Generator
(PMSG)-based wind turbine:
PMSG Grid

~
Renewable energy integration
Power Electronic Converters (PECs) are used to interface the
renewable sources with the grid. PECs are needed to facilitate the
exchange of energy between two (or more) subsystems i.e. Power
Conditioning e.g. :
• Photovoltaic (PV) systems generate DC, hence we need DC/AC
inverters to connect PV to the grid
• Due to the varying nature of the input power (i.e. solar irradiation
and wind speed), in most application, we need to make sure that we
extract the maximum power, which is called maximum power
point tracking (MPPT). This point occurs at a specific operating
point (e.g. a specific voltage or rotor speed). We need PECs to
impose that operating point in order to do MPPT.
Renewable energy integration
• Wind turbines/PV, due to their variable input power, generate
variable frequency/voltage electricity. Hence, PECs are needed to
connect it to a constant frequency/voltage system that is the grid.
PECs can be categorized into:
• DC/AC converters
• Inverters change DC to AC
• Rectifiers change AC to DC
• DC/DC converters
• Buck, output voltage < input voltage
• Boost, output voltage > input voltage
• AC/AC converters
• Indirect: two back to back AC/DC-DC/AC (AC/DC/AC)
• Direct: Matrix Converters
Renewable energy integration
Power Electronic Switches:

• Uncontrollable: No control on its ON and OFF status i.e.

conduction and interruption are determined by the host circuit not an

external control regime. E.g. diodes

• Semi-controllable: E.g. thyristors where a gating signal determines

conduction, but no control on the interruption time (OFF).

• Fully controllable: Both current conduction and interruption can be

controlled. E.g. MOSFET, IGBT, GTO, etc.

Renewable energy integration
Converters classification based on commutation process:
• Line-commutated converter (LCC): whenever the AC side voltage
is reversed the commutation will be initiated. E.g. in a six-pulse
thyristor bridge rectifier, where each switch is turned on using a
gating signal, and turned off when the
AC voltage is revered.

• Forced (self) commuted converter: The process of transferring the

current from one switch to another can be controlled. Hence, we
need fully controllable switches to build the converter.
Renewable energy integration
Converters classification based on terminal voltage and current:

• Current sourced converter (CSC):

The polarity of the DC side current does not change.

L
DC/AC

Hence, the direction of the average power can be controlled by the

polarity of the DC side voltage.
To keep the DC current continuity, a large inductor is connected to the
DC side.
A six pulse thyristor bridge (used in LCC) is a CSC.
Renewable energy integration
• Voltage sourced converter (VSC):
The polarity of the DC side voltage does not change.

C DC/AC

Hence, the direction of the average power can be controlled by the

polarity of the DC side current.
The DC side of VSC is normally connected to a large capacitor.
Renewable energy integration
L
C DC/AC DC/AC

Although forced (self) commuted converter (that uses fully

controllable switches) can be designed to operate as VSC or CSC,
forced commutated CSC is not as common as VSC.

Therefore, in this chapter we will only concentrate on the control of a

forced (self) commuted 3-phaseVSC, which in short we will call it
hereafter a VSC.
The different circuit configuration of a (3-phase) VSC is not also the
subject of this module.
Renewable energy integration
Pcon
The control system produces the Qcon Zf =R+jωL
C Vdc Pdc
V Iabc
modulating signals mabc, which t-abc Cf Vg-abc

are compared with the carrier

Control mabc PWM
waveform to generate the PWM system

signals.

2-level, and 3-level

VSCs can be averagely
modelled as:
Vt − abc = 0.5mabcVdc
2-level, 3-phase VSC
Renewable energy integration
Pcon
Qcon Zf =R+jωL
C Vdc Pdc
V Iabc
t-abc Cf Vg-abc

Control mabc
Vt − abc = 0.5mabcVdc
PWM
system

• Vt-abc can be controlled through controlling Iabc, which is called

current controlled VSC, and is the subject of this chapter.

• Direct voltage control of VSC is also possible (i.e. no current

loop).

• This chapter will focused on the current controlled VSC as it is

more common and provides more flexible control.
Renewable energy integration
Ipv Pcon
ZF CB Grid
Qcon Pg
Ppv Vdc Pdc
Icon Qg ZT ~
Vcon
C Vc

Variable Load
PL, QL
Ipv Pcon
Qcon ZF Pg
Ppv Vdc Pdc
Icon Qg
Vcon
C Vc

The control of a VSC will be explained in the context of PV systems.

However, the principles are applicable to other sources as well.

Most (if not all) of the topics in this chapter are subject of ongoing
research and hence subject to change.
Renewable energy integration
Ipv Pcon
ZF CB Grid
Qcon Pg
Ppv Vdc Pdc
Icon Qg ZT ~
Vcon
C Vc

Variable Load
PL, QL
Ipv Pcon
Qcon ZF Pg
Ppv Vdc Pdc
Icon Qg
Vcon
C Vc

Modes of operation:
• Grid-connected (CB is closed). Normally operates in MPPT
• Islanded (CB is open). Load determines the generation
• Standalone i.e. only one unit
• More than one units. A load sharing scheme is required
Renewable energy integration
Ipv Pcon
ZF CB Grid
Qcon Pg
Ppv Vdc Pdc
Icon Qg ZT ~
Vcon
C Vc

Variable Load
PL, QL
Ipv Pcon
Qcon ZF Pg
Ppv Vdc Pdc
Icon Qg
Vcon
C Vc

Islanding can be planned (e.g. for maintenance) or unplanned (e.g.

due to a fault).
Therefore, conventionally, islanding detection methods are needed to
detect an unplanned islanding event and switch from grid-connected to
islanded mode.
Renewable energy integration
Ipv Pcon
ZF CB Grid
Qcon Pg
Ppv Vdc Pdc
Icon Qg ZT ~
Vcon
C Vc

Variable Load
PL, QL
Ipv Pcon
Qcon ZF Pg
Ppv Vdc Pdc
Icon Qg
Vcon
C Vc

In grid-connected mode, the voltage and frequency are imposed by

the grid. Therefore, the DG units must be synchronised to the grid
frequency (just similar to a synchronous generator which is
synchronised to the grid frequency). This is usually done using a
Phase Locked Loop (PLL) system. PLL generates an output signal
which is in phase with the input signal (more on PLL later on).
Renewable energy integration
Ipv Pcon
ZF CB Grid
Qcon Pg
Ppv Vdc Pdc
Icon Qg ZT ~
Vcon
C Vc

Variable Load
PL, QL
Ipv Pcon
Qcon ZF Pg
Ppv Vdc Pdc
Icon Qg
Vcon
C Vc

In islanded mode, the DG units (operating as a “microgrid”), must be

able to supply (at least) the (critical) loads. Hence, DG units need to
control the voltage and frequency locally:
• Master-slave: One DG (master) imposes the frequency and the
other units (slaves) are synchronised to it. Reliability can be
compromised as the whole system depends on the master.
• Droop control: All DGs share the control of the local network.
Renewable energy integration
The control system can be designed in 3 different reference frame:

• Natural reference frame (NARF), this simply is a 3-phase system

(abc) i.e. one control loop for each phase. Control is difficult since
they are 3 AC waveform to control, hence 3 controllers.

• Stationary reference frame (STRF) also known as αβ frame. The

phasor of abc signals, at any instant, projected in a fixed frame. The
result is still AC but now reduced to 2 waveform.

• Synchronous reference frame (SYRF) also known as dq frame.

The abc signals are projected in a frame that is synchronised to the
abc frequency and is rotating with it. Hence, the projected signals in
dq frame are constant as if they are DC. Thus, the control design is
much simpler.
Renewable energy integration

This chapter will mainly focus on dq rotating frame.

Next, first the αβ frame will be introduced, which leads us to dq frame.
Space Phasor
Considering a balanced 3-phase system:
ua = U cos(t +  0 )
U is the amplitude
 2 
ω is the angular frequency (rad/s) ub = U cos t +  0 −  (1)
 3 
θ0 is the initial phase angle  4 
uc = U cos t +  0 − 
 3 
The space phasor representation is defined as:
2 4
2  jo j  j
u s = ua e + ub e + uc e  (2) 3 3
3 
Substituting (1) into (2) and using cos( ) = 0.5(e j + e − j )
2 4
j j
and (e j 0 + e 3
+e 3
) = 0, gives :
u s = Ue j (t + 0 )
αβ-frame:
u s = Ue j (t + 0 )

I.e. us contains all the information that (1) has.

So we can represent the 3-phase system with us:

u s = u + ju 
u = U cos( ) and u  = U sin( )
d
= (θ = ωt+θ0)
dt

I.e. uα and uβ are time varying sinusoidal functions with amplitude of

U and frequency of ω.
 αβ-frame:
Since us represent a 3-phase system, αβ can be used to represent abc:
This called Clarke transform: abc to αβ:
22 44
= u + juuus == 2 uua eejojo ++uub ee 33 ++uuc ee 33 
2  j j j j
(2)
(2)
s
33  a b c

= U cos( ) and u  = U sin( ) jj −−jj
Substituti ng (1) into (2) and using cos(
Substituting (1) into (2) and using cos( ) = 0.5(e + e ))
) = 0 . 5( e + e
d  2 − 1 2− 1  44u
= u   jj00 jj 33  jj 33a  2

dt   =and
and 3((ee 3++ee 3 ++ee u ))== 00,,givesgives::
 1 − 1   b
u
   uus=0=Ue jj((tt++00))
 u 
Ue
s
 3 3   c 
 
1 0 
 a 
u
u  =  − 1 3  u 
Inverse Clarke transform: αβ to abc :  
 b  2 2  u  
uc   − 1 − 3
 
2 2 
αβ-frame:

3-phase power in αβ-frame:

(Note that v and i are peak phase values)

𝑆 = 1.5 (𝑣)(𝑖)∗ = 1.5(𝑣𝛼 + 𝑗𝑣𝛽 ) × (𝑖𝛼 − 𝑗𝑖𝛽 ) = 𝑃 + 𝑗𝑄

𝑃 = 1.5 𝑣𝛼 𝑖𝛼 + 𝑣𝛽 𝑖𝛽
𝑄 = 1.5 𝑣𝛽 𝑖𝛼 − 𝑣𝛼 𝑖𝛽
dq-frame (rotating frame):

αβ-frame is also known as stationary frame.

It is simply because the frame is fixed and the
phasor is rotating to represent a 3-phase
system.

Hence, in αβ, the signals are sinusoidal function

of time. So, the controller must has higher order
and bandwidth to achieve small steady state error.
Thus, Proportional Resonant (PR) controllers are
used.
dq-frame (rotating frame):
β us
q ω d
ω
uq ud
α

Since the dq-frame also rotating at the same

speed, the signals are DC (at steady state),
which make the control design much
simpler. Hence PID controllers can be used.
dq-frame (rotating frame): β us
q ω d
➢ d is known as direct axis. ω
➢ q is known as quadrature axis. uq ud
➢ Park Transform: αβ to dq α

So dq-frame can be
considered as rotating
αβ-frame.

By choosing  =  dt =t

(i.e. synchronising dq frame to the grid frequency, ud and uq will remain
DC signal.
dq-frame (rotating frame):
β us
q ω d
ω
uq ud
α

It is normally assumed that at t=0 s, the dq frame is aligned with phase-

a axis (θ0=0). This is also known as cosine-based Park
transformation.
dq-frame (rotating frame):
u  cos( ) − sin( ) ud 
Inverse park transform: u  =    
    sin( ) cos( )  uq 

It is also possible to directly transform abc to dq:

 2 4  u 
ud  2  cos( ) cos( − 3 ) cos( − 3 )   
a

u  =    u b 
 q  3 − sin( ) − sin( − 2 ) − sin( − 4 ) u 
 3 3   c

And the inverse:

 
ua   cos( ) − sin( ) 
u  = cos( − 2 ) − sin( − 2 )  ud 
 b  3
  
3  u q 
uc  
cos( − 4 ) − sin( − 4 )
 3 3 
dq-frame (rotating frame):

3-phase power in dq-frame:

(note that v and i are peak phase values):

𝑆 = 1.5(𝑣)(𝑖)∗ = 1.5(𝑉𝑑 + 𝑗𝑉𝑞 ) × (𝐼𝑑 − 𝑗𝐼𝑞 ) = 𝑃 + 𝑗𝑄

𝑃 = 1.5 𝑉𝑑 𝐼𝑑 + 𝑉𝑞 𝐼𝑞
𝑄 = 1.5 𝑉𝑞 𝐼𝑑 − 𝑉𝑑 𝐼𝑞

Note that if Vq=0 (i.e. V=Vd), P and Q are proportional to Id and

Iq, respectively.
Grid-connected VSC Pcon P
Zf =R+jωL Q Grid
in dq frame: C Vdc Pdc
Qcon
~
V Iabc ZT
t-abc Cf Vg-abc

Iabc Control mdq mabc

abc/dq system in PWM Vg-abc θ
dq/abc PLL
Vg-abc dq f
θ θ

For now let’s assume that the DC side of VSC is connected to a constant
DC voltage source.
The control purpose is to independently control the active and reactive
power from the VSC, through controlling its AC current in dq frame.
The PLL provides the grid phase angle in order to synchronise the dq
frame with it.
Grid-connected VSC θ
Vgd
kp ω0
in dq frame: Vg-abc abc/dq Vgq ω θ
0 - ki
0 = 2f 0 in UK f 0 = 50 Hz 1/2π f

There are many different types of PLL. The most common one, which is
shown above, is called: synchronously-reference-frame (SRF)-PLL.
PLL synchronises the VSC with the grid frequency through making
Vgq=0. Why?

Vga = V cos(t )
space phasor:
 2  vs = Ve j (t )
Vgb = V cos t − 
 3 
 4 
Vgc = V cos t − 
 3 
Grid-connected VSC θ
Vgd
kp ω0
in dq frame: Vg-abc abc/dq Vgq ω θ
0 -
0 = 2f 0 in UK f 0 = 50 Hz ki 1/2π f

Assuming that the angle used in abc to dq transform is not synchronised

to grid frequency (θ≠ωt), the space phasor in dq frame is:
Vgd = V cos(t −  )
( )e
j ( t ) − j 
Vgd + jVgq = Ve →
Vgq = V sin(t −  )
If Vgq=0 sin(ωt-θ)=0 ωt=θ
Design of the PLL’s PI controller is beyond the scope of this module.
Since the control plant is nonlinear, the nominal frequency ω0 is added
to the controller output, to reduce the time needed for synchronisation.
Plus, ω will be limited (e.g. ±2 Hz). When θ=2π, the integrator will be
reset to 0 (this is a voltage-controlled-oscillator, VCO)
Grid-connected VSC Pcon P
Zf =R+jωL Q Grid
in dq frame: C Vdc Pdc
Qcon
~
V Iabc ZT
t-abc Cf Vg-abc

Iabc Control mdq mabc

abc/dq system in PWM Vg-abc θ
dq/abc PLL
Vg-abc dq f
θ θ

Pcon  P = 1.5(Vgd I d + Vgq I q ) ⎯⎯ ⎯ ⎯

Vgq = 0 ,Vgd =V
⎯→ Pcon  P = 1.5VI d
Qcon  Q = 1.5(Vgq I d − Vgd I q ) ⎯⎯ ⎯ ⎯
Vgq = 0 ,Vgd =V
⎯→ Qcon  Q = −1.5VI q

Since Vgq=0 →V=Vgd , P and Q are decoupled and

proportional to Id and Iq, respectively.
So, to control P and Q, we need to control Id and Iq:
Current Loop: Pcon P Grid
Zf =R+jωL Q
Qcon
C Vdc Pdc ~
V Iabc ZT
t-abc Cf Vg-abc

Id Compensation
- terms
Id* md
C(s)

Iq* mq
C(s)
-

Iq Compensation
terms

P* and Q* are, respectively, the reference active and reactive power

that the VSC is supposed to generate. This is achieved by forcing Idq
follow their reference Idq*, through using the controller C(s). C(s) is
conventionally a PI controller: ki k p s + ki
C (s) = k p + =
s s
Current Loop:
Id
-
Id* Fd Id
C(s) Gd(s)

Iq* Fq Iq
C(s) Gq(s)
-
ki k p s + ki Iq
C (s) = k p + =
s s

In order to design a proper C(s) i.e. to find appropriate kp and ki, we

need to work out the control plant G(s).
Current Loop: Pcon
Qcon Zf =R+jωL Q
P Grid
C Vdc Pdc ~
V Iabc ZT
t-abc Cf Vg-abc

Id Compensation
- terms
Id* md
C(s)

Iq* mq
C(s)
-

Iq Compensation
terms
di
KVL (in space phasor): vt = Ri + L + v g
dt

Space phasor u in dq is: U dq = (U d + jU q ) = ue − jt → u = U dq e jt

So, the KVL in dq is: d ( I dq e jt )

Vt − dq e jt = RI dq e jt + L + Vg − dq e jt
dt
d ( I dq e jt )
L = Vt − dq e jt − RI dq e jt − Vg − dq e jt
dt
Current Loop:

d ( I dq e jt )
L = Vt − dq e jt − RI dq e jt − Vg − dq e jt →
dt
dI dq j t j t j t j t j t  e − j t
L e + jI dq Le = Vt − dq e − RI dq e − Vg − dq e ⎯⎯⎯→
dt
dI dq U =U + jU
L = − jI dq L + Vt − dq − RI dq − Vg − dq ⎯⎯
dq
⎯d ⎯
⎯q →
dt
 dI d
 L dt = LI q + Vtd − RI d − Vgd

 L dI q = − LI + V − RI − V It is noted that
 dt d tq q gq
• At steady state: ω≈ω0
• There is a cross-link between the 2
equations through LωIdq.
Current Loop:
sLI d + RI d = LI q + Vtd − Vgd
Rearrange and write in Laplace domain: 
sLI q + RI q = − LI d + Vtq − Vgq

Due to the presence of Lω terms, the dynamic of Id and Iq are coupled.

In order to decouple them, a new control input of Fdq is defined:
sLI d + RI d = Fd , Fd = LI q + Vtd − Vgd

sLI q + RI q = Fq , Fq = − LI d + Vtq − Vgq

Thus: Id
Id 1 Id*
-
Fd Id
Gd ( s ) = = C(s) Gd(s)
Fd R + Ls
Iq
Iq 1 Iq* Fq
Gq ( s ) = = -
C(s) Gq(s)
Fq R + Ls
Iq
Current Loop: Id 1
Gd ( s ) = =
Fd R + Ls
These equations describe 2 decoupled, 1st
Iq 1
order, linear systems. Gq ( s ) = =
Fq R + Ls
Idq can be controlled by Fdq. However, Fdq is not the real control input,
but mdq is. The average model of VSC in dq domain is: Vtd = 0.5mdVdc
Vtq = 0.5mqVdc
Thus, from Fdq,
through adding Id Vgd
-
Fd Vtd md
Id*
‘compensation terms’, C(s) /
-
Vt-dq is calculated, Iq Lω mabc
0.5Vdc dq/abc PWM
Id Lω
which will be divided Vtq mq
Iq* Fq /
C(s)
by 0.5Vdc to get mdq: -
Iq Vgq θ
Current Loop:

k p s + ki 1
C (s) = G (s) =
s R + Ls
Obviously, C(s) is the same for both d and q loop since Gd(s)=Gq(s).

The compensation terms will be ignored in designing the controller C(s).

There are several methods for designing C(s) such as: Root Locus and
Frequency Analysis. Here we consider two simple ways. :
Pole Cancelation: The open loop transfer function (OLTF) is:
 k 
kps +  i
 k 
OLTF = C ( s )  G ( s ) =  
p

 R
Ls s + 
 L
Current Loop:
 k 
kps +  i
 k 
OLTF = C ( s )  G ( s ) =  
p

 R
Ls s + 
 L
Since the plant has pole at s=-R/L, which is relatively close to the
origin, one method is to use the zero of the controller to cancel it i.e.
ki R
=
kp L
Thus, the closed loop transfer function (CLTF) will be:
kp
I C (s)  G (s) Ls 1 L
CLTF = * = = = where ti =
I 1 + C (s)  G (s) 1 + k p ti s + 1 kp
Ls
Hence, the bandwidth of the response of I to I* (i.e. fi) is determined by
the 1st order the low-pass filter: 1
i = 2f i =
ti
Current Loop:
As a general guideline, the controller bandwidth should be at least 10
times slower than the switching frequency:
There are other criteria, e.g.: 1  sw = 2f sw
i = 2f i = 
ti 10

Choosing ti from the above criterion,

L Rk p
kp = and ki =
ti L

The other method to design the current controller is to use the

characteristic equation:
Current Loop:

kp
I C (s)  G (s) 1 k p s + ki L 1
CLTF = * = = Ls =C ( s ) = where t = G ( s ) =
I 1 + C (s)  G (s) 1 + k p ti s + 1 s i
kp R + Ls
Ls
Characteristic equation is the denominator of CLTF=1+OLTF:

k p s + ki Ls 2 + ( R + k p ) s + ki (R + k p ) ki
Ch. Eq. = 1 + = =0→s +
2
s+ =0
s ( R + Ls ) s ( R + Ls ) L L

Comparing it with the standard 2nd order system:

 2 ki

 n = → k i = L  n
2

L
s + 2n s + n
2
2
=0→
2 = ( R + k p ) → k = 2 L − R
 n
L
p n
Current Loop:

 2 ki

 n = → k i = L  n
2

L
s 2 + 2n s + n
2
=0→
2 = ( R + k p ) → k = 2 L − R
 n
L
p n

ωn is the natural frequency of the system which here is approximated

the bandwidth of the controller. Thus, the same criteria exists for
choosing it:
 sw = 2f sw
n = i = 2f i 
10
 is the damping ratio and is normally chosen between:
=0: transient response does not die out
0<  <1: under-damped: oscillatory (2 complex conjugates CL poles)
=1: critically damped (the 2 CL poles are real and equal)
>1: over-damped (the 2 CL poles are real – can be sluggish)
DC-link voltage control:
For a given solar irradiation, there is only
one Vpv (called optimum voltage) at which
Ppv is a maximum. Hence, to extract the
maximum power we should control Vpv.
This is called maximum power point
tracking (MPPT).
There are different MPPT algorithms aiming to provide the optimum
voltage as a reference Vpv* for a voltage control loop, as solar irradiation
varies: DC/AC
Ipv
ZF Grid
Vpv
~
C

Ipv Vpv* Id*

MPPT Voltage
Loop mabc
Vpv Current PMW
Loops
Q* Q Iq*
Loop
 Ppv Pdc Pcon P
DC-link
voltage Pcap+Ploss
Ipv Pcon P
control: Qcon Zf =R+jωL Q Grid
Ppv C Vpv Pdc ~
V Iabc ZT
t-abc Cf Vg-abc
Ipv Vpv* Voltage Id*
MPPT mabc
Vpv Loop Current PMW
Loops
Q* Q Iq*
Loop

The purpose of voltage loop is to transfer Ppv to AC-side. This is

achieved by keeping Vpv constant (it is noted that as solar irradiation
varies, Vpv* is, transiently, changed by the MPPT algorithm to extract
the maximum power. The voltage loop is supposed to Vpv=Vpv* to do
MPPT and transfer Ppv to the AC-side).
Assuming zero loss on the DC link: Ppv=Pcap + Pdc.
𝑑𝑉𝑝𝑣 𝑑(𝑉𝑝𝑣 2 )
The capacitor power: 𝑃𝑐𝑎𝑝 = 𝑉𝑝𝑣 𝐶 = 0.5𝐶
𝑑𝑡 𝑑𝑡
 Ppv Pdc Pcon P
DC-link
voltage Pcap+Ploss
Ipv Pcon P
control: Qcon Zf =R+jωL Q Grid
Ppv C Vpv Pdc ~
V Iabc ZT
t-abc Cf Vg-abc
Ipv Vpv* Voltage Id*
MPPT mabc
Vpv Loop Current PMW
Loops
Q* Q Iq*
Loop

Assuming a lossless VSC and ignoring the filter losses:

Pdc≈Pcon≈P≈1.5VgdId (note that Vgq≈0 → Vgd ≈V), thus:
2
dV pv
Ppv = 0.5C + 1.5Vgd I d
dt
2
dV pv Ppv 1.5Vgd
= − Id
dt 0.5C 0.5C
DC-link voltage Id
Fv - Id Fv
control: (V *) pv
2 - Id* Fd
Gd(s)
Vpv2
Cv(s) C(s) Gv(s)
Compensation CLTFi 1 Compensation
terms terms
2
dV pv
=
1
(Ppv − 1.5Vgd I d ) ⎯Laplace 2
⎯⎯→V pv =
1
(Ppv − 1.5Vgd I d ) = 1 Fv
dt 0.5C 0.5Cs 0.5Cs
Applying Laplace transform, the voltage loop plant will be:
2
V pv
1
Gv = = , Fv = Ppv − 1.5Vgd I d
Fv 0.5Cs
• Since C is constant, the plant is an integrator.
• The current (inner) loop must be at least 10 times faster than the
voltage (outer) loop. Hence, in designing the controller Cv(s) we can
assume that: CLTFi=1 i.e. Id*=Id → current loop can be ignored
• As always, the compensation terms are ignored in designing.
DC-link Fv = Ppv − 1.5Vgd I d (Vpv*)2
- Fv Vpv2
Cv(s) Gv(s)
voltage
k pv s + kiv 1
control: Gv =
Cv = 0.5Cs
V pv
2
Cv  Gv s
CLTF = =
V pv *2 1 + Cv  Gv
Different methods can be used to design Cv. Here we use
characteristics equation:
Characteristic equation is the denominator of CLTF=1+CvGv:
k pv s + kiv 0.5Cs 2 + k pv s + kiv k pv
kiv
Ch. Eq. = 1 + 2
= =0→s + 2
s+ =0
0.5Cs 0.5Cs 2 0.5C 0.5C

 2 kiv
n = 0.5C → kiv = 0.5Cn
2

s 2 + 2n s + n
2
=0→
2 = k pv → k =  C
 n
0.5C
pv n
DC-link Fv = Ppv − 1.5Vgd I d (Vpv*)2
- Fv Vpv2
Cv(s) Gv(s)
voltage
k pv s + kiv 1
control: Gv =
Cv = 0.5Cs
s
 2 kiv

 n = → k iv = 0 . 5C  n
2

0.5C
s 2 + 2n s + n
2
=0→
2 = k pv → k =  C
 n
0.5C
pv n

ωn is the natural frequency of the system which here is approximated

the bandwidth of the voltage control loop ωv, which should be around
10 times slower than the current loop ωi:
i = 2f i
n = v = 2f v 
10

 is the damping ratio and is normally chosen between: 0.6-0.9

DC-link voltage control:
- Fv
(Vpv*)2 Cv(s) Vpv2
Gv(s)

Note that Fv is not the actual control input, Id is. So compensation terms
are considered to calculate the current reference:
Id =Id * * Ppv − Fv
Fv = Ppv − 1.5Vgd I d ⎯⎯⎯→ I d =
1.5Vgd

Id Vgd
Vpv* (Vpv*)2 Fv - Id* - Vtd md
MPPT ^2 Cv(s) Fd /
- / C(s)
Vpv Vpv2 Ppv
-
^2 Iq Lω mabc
1.5Vgd 0.5Vdc dq/ PWM
Id Lω abc
Iq* Fq Vtq mq
-Q* /
/ C(s)
-
Iq Vgq
θ
Current loop
Reactive power control: Id Vgd
2
Vpv* (Vpv*) Fv - Id* - Vtd md
MPPT ^2 Cv(s) Fd /
- / C(s)
Vpv Vpv2 Ppv
-
^2 Iq Lω mabc
1.5Vgd 0.5Vdc dq/ PWM
Id Lω abc
Iq* Fq Vtq mq
For reactive power control -Q*
/ C(s) /

we know that: -
Iq Vgq
−Q* θ
Q = −1.5Vgd I q → I q * = Current loop
1.5Vgd
• Controlling Q as above, is not very precise as it is an open loop
control (no feedback). Thus normally a power loop is also designed.
• Normally Q*=0 in order to have a unity PF (PF=1).
• It is also possible to control Iq in order to regulate the PF or the
converter voltage. This is similar to AVR systems (and excitation
current) in synchronous generators.
• In modern networks, the DGs are required to provide ancillary
services such as voltage and frequency support. This is known as
Active Network Management.
 DC/AC
DC-link Ipv
Grid
ZF
voltage Vpv DC/DC Vdc
~
C
control:
Ipv Vpv* Voltage
MPPT
Vpv Loop

Vdc* Voltage Id*

Loop mabc
Current PMW
Loops
Q* Q Iq*
Loop

In a 2-stage PV system the DC/DC converter controls Vpv to perform

MPPT, while the DC-link voltage of the DC/AC converter (Vdc) is
controlled by another voltage loop through Id.

In a wind turbine, the MPPT is done through controlling the rotor

speed at the “optimum speed” where the power is maximum for a
given wind speed (and pitch angle).
Islanded operation (off-grid):
Ipv Pcon
ZF CB Grid
Qcon Pg
Ppv Vdc Pdc
Icon Qg ZT ~
Vcon
C Vc

Ipv Pcon PL, QL

Qcon ZF Pg
Ppv Vdc Pdc
Icon Qg
Vcon
C Vc

Islanding can be planned (e.g. for maintenance) or unplanned (e.g. due

to a fault). Therefore, conventionally, islanding detection methods are
needed to detect an unplanned islanding event and switch from grid-
connected to islanded mode.

• In islanded (CB is open) mode, the load determines the generation

• Standalone i.e. only one unit
• More than one units. A load sharing scheme is required
Islanded operation:

In islanded mode, the DG units (operating as a “microgrid”), must be

able to supply (at least) the (critical) loads. Hence, DG units need to
control the voltage V and frequency f locally:
• Master-slave: One DG (master) imposes the frequency and the other
units (slaves) are synchronised to it. Not very secure as the whole
system depends on the master.
• Droop control: All DGs share the control of the control of V and f
through sharing the loads PL and QL.
 Ipv Pcon
Islanded Qcon R+jωL
P
Q
Ppv C Vpv Pdc PL, QL
operation: V Iabc
Cf
IL-abc
t-abc Qc Vg-abc

For start let’s consider only one unit. The DG unit is supposed to supply
the load i.e. at steady state P=PL and Q=QL. Therefore:

The control objective is to regulate the amplitude

and frequency of load voltage Vg-abc.

It is assumed that at any instant of time:

• PL≤ maximum power point (Pmpp) of the solar irradiation at the time
• Load apparent power S L = PL 2 + QL 2  S rated of the VSC

In practical systems energy storage (e.g. a battery) will be used to store

the extra energy when PL<Pmpp and to supply the shortage of energy
when PL>Pmpp. This is beyond the scope of this module.
 Ipv Pcon
Islanded P
Qcon R+jωL Q PL, QL
Ppv C Vpv Pdc
operation: Iabc IL-abc
V t-abc Cf Qc Vg-abc

IL-abc Iabc Control mdq mabc

abc/dq system in dq/abc PWM
Vg-abc dq

ω=ω0 θ θ
θ=ωt

The control will be done in dq frame. Therefore, the measured currents

and voltage are transferred to dq, and then the control output, which is
the modulating signal mdq will be transferred back to abc frame.

Since there is no grid, there is no need to a PLL to synchronise the DG

with the grid. However, the dq frame must still rotate at the fundamental
frequency ω0=2πf0 (f0=50 Hz in UK). Thus, a Voltage-Controlled
Oscillator (VCS), which is simply a resetable integrator (limited 0-2π) is
used to impose the rotation of the dq frame:
 =  0 dt = 0t = t
Islanded
Ipv Pcon
operation: Qcon R+jωL
P
Q
Ppv C Vpv Pdc PL, QL
Iabc IL-abc
V t-abc Cf Qc Vg-abc

IL-abc Iabc Control mdq mabc

abc/dq system in dq/abc PWM
Vg-abc dq

ω=ω0 θ θ
θ=ωt
Islanded
operation:

The internal current loop is identical to the grid-connected mode.

By making the internal current loop at least 10 times faster than the
external voltage loop → CLTFi≈1 → the current loop can be ignored
in designing the voltage loop i.e. Idq*≈Idq . Therefore:
 Ipv Pcon
Islanded Qcon R+jωL
P
Q
Ppv C Vpv Pdc PL, QL
operation: V Iabc
Cf
IL-abc
t-abc Qc Vg-abc

IL-abc Iabc Control mdq mabc

abc/dq system in dq/abc PWM
Vg-abc dq

ω=ω0 θ θ
θ=ωt

Writing KCL (in space phasor) for the node of filter capacitor Cf :
dv g
ic = C f = i − iL
dt

Generally, space phasor u in dq is: U dq = (U d + jU q ) = ue − jt → u = U dq e jt

KCL in dq frame:

Cf
(
d Vg − dq e jt )= I e j t
− I e j t
dq L − dq
dt
Islanded operation:

d (Vg − dq e jt )
Cf = I dq e jt − LL − dq e jt → Apply Product Rule on Line 1 to get Line 2
dt
dVg − dq j t j t j t j t  e − j t
Cf e + jVg − dq C f e = I dq e − I L − dq e ⎯⎯⎯→
dt
Multiply with e^-jwt to get simplify line 2 to line 3.
dVg − dq U =U + jU
Cf = − jVg − dq C f  + I dq − I L − dq ⎯⎯
dq
⎯d ⎯
⎯q →
dt
 dVgd
C f = C f Vgq + I d − I Ld
dt Real Part of the above which relates to d -ive absent due to j2=-1

C dVgq = −C V + I − I
 f dt f gd q Lq It is noted that:
Imaginary part is related to q for above equation • ω=ω0
• There is a cross-link between the 2
equations through Cf ωVg-dq.
Islanded operation:
sC f Vgd = C f Vgq + I d − I Ld
write in Laplace domain: 
sC f Vgq = −C f Vgd + I q − I Lq

Due to the presence of Cf ω terms, the dynamic of Vgd and Vgq are
coupled. In order to decouple them, a new control input of Ndq is
defined:
sC f Vgd = N d , N d = C f Vgq + I d − I Ld

sC f Vgq = N q , N q = −C f Vgd + I q − I Lq
Thus:
Vgd 1 Vgd* -
Nd Vgd
Gvd ( s ) = = Cg(s) Gvd(s)
Nd C f s
Vgq* Nq Vgq
Vgq 1 Cg(s) Gvq(s)
Gvq ( s ) = = -
Nq C f s
Islanded
operation:

sC f Vgd = C f Vgq + I d − I Ld


sC f Vgq = −C f Vgd + I q − I Lq

So, it is possible to use Id and Iq to regulate Vgd and Vgq, respectively.

The current loop is designed exactly the same as grid-connected mode.
Assuming that the current loop is much faster than outer Vg-dq loop, in
designing the Vg-dq loop we can consider CLTFi≈1→ Idq*=Idq:

sC f Vgd = N d , N d = C f Vgq + I d − I Ld → I d = N d − C f Vgq + I Ld

*


, N q = −C f Vgd + I q − I Lq → I q = N q + C f Vgd + I Lq
*
sC f Vgq = N q
Islanded operation:
sC f Vgd = N d , N d = C f Vgq + I d − I Ld → I d = N d − C f Vgq + I Ld
*


, N q = −C f Vgd + I q − I Lq → I q = N q + C f Vgd + I Lq
*
sC f Vgq = N q

Adding the compensation terms:

Vgd ILd Id Vgd

- Nd - Vtd md
Vgd* Id* Fd /
Cg(s) C(s)
- -
Vgq Cf ω Iq Lω
0.5Vdc dq/ mabc
PWM
Vgd Cf ω Id Lω abc

Fq Vtq mq
Vgq* Nq Iq* C(s) /
Cg(s)
- -
Vgq ILq Iq Vgq
θ
Current loop
Islanded operation:
Vgd ILd Id Vgd
Vgd*=1 - Nd Id*
-
Fd Vtd md
Cg(s) C(s) /
pu
- -
Vgq Cf ω Iq Lω
0.5Vdc dq/ mabc
PWM
Vgd Cf ω Id Lω abc

Fq Vtq mq
Nq Iq* /
Vgq*=0 Cg(s) C(s)
- -
Vgq ILq Iq Vgq
θ
Current loop
2 2
| Vg |= Vgd + Vgq  1 pu

The amplitude of Vg must be controlled at around 1 pu. A common

approach is to set Vgq*=0 (similar to a SRF-PLL) and Vgd* =1 pu.
Some literature propose to use a frequency control loop to set Vgq*.
Islanded operation:
Different methods can be used to
𝑘𝑝𝑣 𝑠 + 𝑘𝑖𝑣 1
design Cg. Here we use 𝐶𝑔 = 𝐺𝑣 =
𝑠 𝑠𝐶𝑓
characteristics equation:

Characteristic equation is the denominator of CLTF=1+CgGv=0:

 Ipv Pcon-1 P1
Droop control: Qcon-1 R+jωL Q1
Ppv C Vpv Pdc
I1abc IL1-abc
V t-abc Cf Qc1
PL, QL
Ipv Pcon-2 P2
Qcon-2 R+jωL Q2
Ppv C Vpv Pdc
I2abc IL2-abc
V t-abc Cf Qc2
Vg-abc

If there are more than one unit in islanded mode, a load sharing
algorithm is required.
• Master-slave: One DG (master) imposes the frequency and the other
units (slaves) are synchronised to it. Reliability issues as the whole
system depends on the master.
• Droop control: All DGs share of the control of V and f through
sharing the loads PL and QL. It may suffer from poor transient
response and stability issues.
 Ipv Pcon-1 P1
Droop control: Qcon-1 R+jωL Q1
Ppv C Vpv Pdc
I1abc IL1-abc
V t-abc Cf Qc1
PL, QL
Ipv Pcon-2 P2
Qcon-2 R+jωL Q2
Ppv C Vpv Pdc
I2abc IL2-abc
V t-abc Cf Qc2
Vg-abc

In “conventional” droop control (which is the subject of this section),

the load will be shared proportional to the rating of each unit:
P1 Q1 S rated −1
= =
P2 Q2 S rated − 2

It is assumed that at any instant:

• P1≤ Pmpp-1 and P2≤ Pmpp-2 of the solar irradiation at the time
• Each unit’s apparent power S1  S rated −1 and S 2  S rated − 2
Droop control: 0+jωL
It was shown in Power System module (EG-
342/EGLM07) that assuming R≈0 i.e. a Vt-abc Vg-abc
mainly inductive system, the active and
reactive power sent from Vt:
VtVg
P= sin 
X Where, θ is the phase angle between Vt and Vg.
Vt (Vt − Vg )
Q= cos 
X
• Since dθ/dt=ω,frequency can be controlled by P.
• Since cosθ≈1, voltage can be controlled by Q.
• (It can be shown that in a mainly resistive network, voltage is
controlled by P and frequency is controlled by Q).
Droop control: Here, we consider inductive network:
 ω* V*
 = 0 − mP , m =
*

Pmax ω0 V0
Δω ΔV
V
V = V0 − nQ , n =
*
P
Qmax Pmax Qmax Q

• P and Q are the measured average local powers of each unit

• V0 and ω0 are the nominal values (1 pu) of voltage and frequency
• V* and ω* are the reference voltage and frequency at P and Q
• m and n are called droop gains
• ΔV and Δω are the allowed voltage and frequency deviations (5%
and 2%)
• Pmax and Qmax are maximum active/reactive powers that a unit can
generate/exchange (normally set at their rated apparent power Srated)
Droop control:
Since ΔV and Δω are the same for all units, we will have:

P1m1 = P2 m2 = ... = 
Q1n1 = Q2 n2 = ... = V

This indicates that P and Q are shared in proportional to the units’

ratings:
ω0
V0
ω* V*

P2 P Q
P1 Q1 Q2
Droop control:
Therefore, the complete control system for each unit is:
Vgd ILd Id Vgd θ
*
ω θ Vgd*=V* - Nd - Vtd md
P m Id* Fd /
Cg(s) C(s)
- Vαβ *=V*ejθ - -
ω0
Vgq Cf ω Iq Lω
(this 0.5Vdc dq/ mabc
makes: Vgd Cf ω Id Lω abc
*
Q V Vgd*=V* Vtq mq
n Vgq*=0 Nq Iq* Fq /
- Vgq*=0) Cg(s) C(s)
V0 - -
Vgq ILq Iq Vgq
Droop
Voltage loop Current loop

𝑉𝑔−𝑎𝑙𝑓𝑎 ∗ = 𝑉 ∗ cos 𝜃
 Note that P and Q are the
 = 0 − mP , m =
*
𝑉𝑔−𝑏𝑒𝑡𝑎 ∗ = 𝑉 ∗ sin 𝜃
Pmax
Using αβ to dq average active and reactive
V
V = V0 − nQ , n =
* transform will give:
Qmax Vgd*=V* power, measured after the
and Vgq*=0
filter.
 Pcon-1 P1
Qcon-1 R+jωL Q1
θ1 VDC1
I1 IL1
Cf Vg1
dq/abc
mabc-1
Q1 P1
Vg1-d*=V1*
mdq-1 Voltage
Current I1-dq* Droop
loop
loop Vg1-q* =0 1/S PL, QL
ω1*
θ1
Vg1-dq
Vg1-dq Vg1
VDC1
IL1-dq abc/dq IL1
I1-dq I1

Pcon-2 P2
θ2 Qcon-2 R+jωL Q2
VDC2
I2 IL2
Cf Vg2
dq/abc
mabc-2
Q2 P2
Vg2-d*=V2*
mdq-2 Voltage
Current I2-dq* Droop
loop
loop Vg2-q* =0 1/S
ω2*
θ2
Vg2-dq VPCC
Vg2-dq Vg2
VDC2
IL2-dq abc/dq IL2
I2-dq I2